CN113498071A - Method, apparatus and program for predicting future quality of service of a wireless communication link - Google Patents

Abstract

The present invention relates to a method, apparatus and program for predicting the future quality of service of a wireless communication link, and in particular to an apparatus, method and computer program for predicting a wireless communication link based on a predicted future environment model using time series prediction prediction Road future service quality. The method includes determining a plurality of environmental models of one or more active transceivers in the environment of the mobile transceiver at a plurality of points in time. The method includes determining a predicted future environment model for one or more active transceivers at future points in time using time series predictions of the plurality of environment models. The method includes using a machine learning model to predict future quality of service of the wireless communication link at future points in time. Machine learning models are trained to provide information about the predicted quality of service for a given environment model. The predicted future environment model is used as the input to the machine learning model.

Description

Method, apparatus and program for predicting future quality of service of a wireless communication link

technical field

The present invention relates to an apparatus, method and computer program for predicting future quality of service of a wireless communication link based on a predicted future environment model predicted using time series forecasting.

Background technique

Communication between mobile transceivers is an area of research and development. For example, in vehicle applications, research is underway to improve both the performance and predictability of wireless communication between vehicles in changing environments. For example, in the context of collaborative driving, prediction of the future quality of service (QoS) of a wireless communication link between two vehicles improves the functionality of vehicle applications as quality of service (QoS) conditions change. In practice, when no predictive QoS (PQoS) is provided, the application may only react to changes and thus may be limited to the lower bound performance of the communication system. PQoS systems can operate on vehicles, communication nodes, using radio access technologies (RATs) such as LTE-V (Long Term Evolution for Vehicle Communications) or 5G-V2X (for vehicle-to-anywhere) in its standalone mode The fifth generation mobile communication system for transaction communication) or IEEE 802.11p (standard of the American Institute of Electrical and Electronics Engineers). Combinations of these techniques can also be applied to multi-RAT systems. In such a PQoS system, vehicles can exchange information about the communication surroundings in order to provide PQoS.

In the literature, channel models (semi-random, such as spatial channel models (SCM); and deterministic, such as ray tracing) provide estimates of path loss and interference from other communication nodes. Statistical models can provide ideas about some kind of mapping between surrounding vehicles and quality of service. An example is shown in "Performance and Reliability of DSRC Vehicular Safety Communication: A Formal Analysis and IEEE 802.11 pVANets: Experimental evaluation of packet inter-reception time provide such model" by Ma, Chen and Refai. The paper "Prediction of Packet Inter Reception Time for Platooning using Conditional ExponentialDistribution" by Jornod, El Assad, Kwoczek and Kürner provides a statistical link between ambient density and reception time between packets. The paper also shows a way to divide the surrounding environment into circular regions to represent the traffic density around the transmitter. It uses the distance between transceivers to estimate the QoS of the link.

SUMMARY OF THE INVENTION

There may be a need for an improved concept for predicting the quality of service of wireless communication links between vehicles.

Embodiments are based on the discovery that previous methods for predicting quality of service have focused on providing predictions for a single point in time, rather than using the gradual development of the environment for tracking a mobile transceiver and thus tracking a mobile transceiver and another mobile transceiver. An evolving prediction method for the quality of service between wireless communication links. In an embodiment of the present disclosure, future environmental models of active transceivers around the mobile transceiver are predicted by performing a time series forecast based on previously generated environmental models. Based on an environmental model predicted for a future point in time (or points in time), a machine learning model is used to determine the quality of service at one or more points in time in the future. Thus, by using time series predictions of the environment model, the development of the quality of service of the wireless communication link affected by changes in the environment is predicted.

Embodiments provide a method for predicting future quality of service of a wireless communication link between a mobile transceiver and another mobile transceiver. The method includes determining multiple environment models of one or more active transceivers in the environment of the mobile transceiver at multiple points in time. The method includes determining a predicted future environment model for one or more active transceivers at future points in time using time series forecasts for the plurality of environment models. The method includes using a machine learning model to predict future quality of service of the wireless communication link at future points in time. The machine learning model is trained to provide information about the predicted quality of service given the environmental model. The predicted future environment model is used as the input to the machine learning model. By performing time series predictions, the radio environment of the mobile transceiver is modeled at future points in time. This predicted future environment model can in turn be used to predict the future quality of service of the wireless link via a machine learning model.

In various embodiments, time series forecasting is performed based on a statistical fitting function or based on a temporal autocorrelation function. This statistics-based approach has low computational overhead.

Alternatively, another machine learning model can be used to perform time series forecasting. Machine learning-based time series forecasting may be useful in scenarios with a large number of correlated features and under high computational cost.

For example, a time series forecast may be determined such that the progress of the environmental model towards a predicted future environmental model is predicted. In other words, data from multiple environmental models can be extrapolated to predicted future environmental models.

For example, time series forecasting can produce predicted future environmental models. Predict future service quality based on a predicted future environment model. In other words, time series forecasting may not be applied to the quality of service itself, but to an environmental model that underlies predictions of future quality of service.

In various embodiments, the future quality of service of the wireless communication link is predicted for at least two future points in time. For example, the predicted future environment model may be predicted for at least two future points in time, and then used to predict future quality of service for at least two future points in time.

For example, by determining a predicted future environment model for one or more active transceivers at at least two future points in time, and using the predicted future environment model for one or more active transceivers at at least two future points in time as The input of the machine learning model can predict the future quality of service of the wireless communication link for at least two future time points. Therefore, predictions of future service quality can be provided on a timeline of future points in time.

In an embodiment, multiple environmental models for one or more active transceivers are determined at multiple points in time. The method may include determining the quality of service of the wireless communication link at multiple points in time. The method may include training the machine learning model using the plurality of environment models at the plurality of time points as training inputs and the quality of service of the wireless communication link at the corresponding plurality of time points as training outputs of the training of the machine learning model. Thus, the machine learning model can be trained on the environment model generated by and/or for the mobile transceiver.

For example, a machine learning model can be trained to implement a regression algorithm. Regression-based machine learning algorithms can be used to determine values (within a certain range), such as future quality of service.

In various embodiments, a machine learning model may be trained to provide a probability distribution of predicted quality of service given an environmental model. Doing so avoids the common quality of service forecasts that skew the forecasts.

In some embodiments, one or more active transceivers are placed on a grid within the environment model. The grid may include multiple contiguous cells. One or more active transceivers may be clustered in each cell within the grid. For example, the grid can be used to facilitate maintenance of environmental models and limit the number of inputs to machine learning models.

For example, the grid may be a circular grid. Transmissions by other active transceivers affect the wireless communication link based on their distance, which can be modeled by a circular grid.

Predicting quality of service may relate to at least one of inter-packet reception time, packet error rate, delay, and data rate. These are the quality of service characteristics that can be predicted using the machine learning model described above.

Embodiments of the present disclosure also provide a computer program having program code for performing the above-described method when the computer program is executed on a computer, processor or programmable hardware component.

Embodiments of the present disclosure also provide an apparatus for predicting future quality of service of a wireless communication link between a mobile transceiver and another mobile transceiver. The apparatus includes one or more interfaces for communicating in a mobile communication system. The apparatus includes a control module configured to perform the above-described method.

Description of drawings

Some other features or aspects will be described, by way of example only, using the following non-limiting embodiments of an apparatus or method or computer program or computer program product and with reference to the accompanying drawings, in which:

Figures 1a and 1b illustrate a flowchart of an embodiment of a method for predicting future quality of service of a wireless communication link between a mobile transceiver and another mobile transceiver;

Figure 1c shows a schematic diagram of an apparatus for predicting future quality of service of a wireless communication link between a mobile transceiver and another mobile transceiver, and a mobile transceiver (such as a vehicle) including the apparatus;

Figure 1d shows a flowchart of an embodiment of a method for training a machine learning model;

Figure 2 shows a table of the development of quality of service attributes related to the environment model over time;

Figures 3a to 3d show schematic diagrams related to the training of machine learning models.

Detailed ways

Various exemplary embodiments will now be described more fully with reference to the accompanying drawings, in which some exemplary embodiments are shown. In the drawings, the thickness of lines, layers or regions may be exaggerated for clarity. Optional components can be represented by broken lines, broken lines, or dashed-dotted lines.

Thus, while the exemplary embodiments are capable of various modifications and alternative forms, embodiments thereof are shown by way of example in the drawings and will herein be described in detail. It should be understood, however, that there is no intention to limit exemplary embodiments to the particular forms disclosed, but on the contrary, exemplary embodiments are to cover all modifications, equivalents, and alternatives falling within the scope of the invention. Throughout the description of the drawings, the same numerals refer to the same or similar elements.

As used herein, the term "or" refers to a non-exclusive "or" unless stated otherwise (eg, "otherwise" or "or in the alternative"). Furthermore, as used herein, terms used to describe a relationship between elements should be construed broadly to include the direct relationship or the presence of intervening elements, unless stated otherwise. For example, when an element is referred to as being "connected" or "coupled" to another element, it can be directly connected or coupled to the other element or intervening elements may be present. In contrast, when an element is referred to as being "directly connected" or "directly coupled" to another element, there are no intervening elements present. Similarly, words such as "between", "adjacent", etc. should be interpreted in a similar fashion.

The terminology used herein is for the purpose of describing particular embodiments only and is not intended to be limiting of the exemplary embodiments. As used herein, the singular forms "a," "an," and "the" are intended to include the plural forms as well, unless the context clearly dictates otherwise. It will also be understood that the terms "comprising", "comprising" or "containing" when used herein designate the presence of stated features, integers, steps, operations, elements or components, but do not exclude one or more other features , an integer, a step, an operation, an element, a component, or the presence or addition of a combination thereof.

Unless otherwise defined, all terms (including technical and scientific terms) used herein have the same meaning as commonly understood by one of ordinary skill in the art to which example embodiments belong. It is also to be understood that terms, such as those defined in commonly used dictionaries, should be construed to have meanings consistent with their meanings in the context of the relevant art, and are not intended to be idealized unless explicitly so defined herein Or in an overly formal sense to explain.

Figures 1a and 1b show a flow diagram of an embodiment of a method for predicting future quality of service of a wireless communication link between a mobile transceiver 100 and another mobile transceiver 102. The method includes determining 110 a plurality of environmental models of one or more active transceivers 104 in the environment of the mobile transceiver at a plurality of points in time. The method includes determining 120 a predicted future environment model for one or more active transceivers at future points in time using 125 the time series predictions of the plurality of environment models. The method includes predicting 130 future quality of service of the wireless communication link at future points in time using a machine learning model. Machine learning models are trained to provide information about the predicted quality of service for a given environment model. The predicted future environment model is used as input to the machine learning model 135.

Figure 1c shows a schematic diagram of a corresponding apparatus 10 for predicting the future quality of service of a wireless communication link between a mobile transceiver 100 and another mobile transceiver 102 and a mobile transceiver 100 (such as a vehicle) comprising the apparatus 10 . The device includes one or more interfaces 12 for communicating in a mobile communication system. The apparatus includes a control module 14 configured to perform at least one of the methods shown in Figures 1a, 1b and/or 1d. In general, control module 14 may, for example, incorporate one or more interfaces 12 to provide the functionality of device 10.

The following description refers to both the method of Figures Ia and/or Ib and the apparatus of Figure Ic. Features described in connection with the method of Figure 1a and/or Figure 1b are equally applicable to the device of Figure 1c.

Embodiments of the present disclosure relate to methods, apparatus and computer programs for predicting future quality of service of a wireless communication link between a mobile transceiver and another mobile transceiver. In order to predict future service availability and QoS, it may be beneficial to have a good understanding of the radio environment. Typical characteristics of a radio environment may be path loss, interference conditions, system loading, number of carrier frequencies, multiple radio access technologies (RATs), etc. The more detailed the radio environment is modeled, the greater the amount of information required for its classification.

As described above, the future QoS of the wireless communication link is determined. In this case, the prefix "future" means predicting the QoS at a point in time in the future. To accomplish this, the presence (and thus activity) of one or more active transceivers can be predicted for the future (eg, using the trajectory of the one or more active transceivers), and the prediction of future QoS can be based on a prediction of one or more active transceivers. Forecast of future activity for multiple active transceivers. The predicted QoS may include one or more aspects, such as predicted minimum, average and/or maximum data transmission rates, minimum, average and/or maximum packet or bit error rates, two successful transmissions of packets on the wireless link Minimum, average or maximum time between (eg, also expressed as inter-packet reception time PIR), minimum, average or maximum delay, etc. In other words, predicting quality of service may relate to at least one of inter-packet reception time, packet error rate, delay, and data rate of the wireless communication link. Generally, the predicted QoS may indicate the expected performance and/or expected reliability of the wireless link.

In an embodiment, the future QoS of the wireless link may be predicted by the mobile transceiver, i.e. by the receiver of the wireless transmission on the wireless link. Thus, the method may be performed by a mobile transceiver, and/or the mobile transceiver may include the apparatus of Figure 1c.

The method includes determining 110 a plurality of environmental models of one or more active transceivers 104 in the environment of the mobile transceiver at a plurality of points in time. Generally, an environment model of one or more active transceivers may model the environment of a mobile transceiver for the presence and/or transmission activity of the one or more active transceivers. For example, the environment model may include and/or represent the location of one or more active transceivers in the environment of the mobile transceiver. In various embodiments, the environment model may be limited to a predetermined range around the mobile transceiver, such as a predetermined circular distance, or a predetermined size according to a grid. For example, one or more active transceivers can be placed on a grid within the environment model. The grid may include multiple adjacent cells. One or more active transceivers may be clustered in each cell within the grid. In other words, the location and/or distance of one or more mobile transceivers may be represented by the cells of the grid in which they are placed.

For example, in some cases, the grid may be a two-dimensional rectangular grid. In other words, the cells of the grid may have a rectangular shape. Additionally, each cell of the grid may have (substantially) the same size/dimension.

Alternatively, the grid may be a circular grid, such as a one-dimensional circular grid. In a one-dimensional circular grid, the cells of the grid are formed along one dimension (i.e., the distance from the center) such that multiple concentric circles form the grid, and the spaces between the circles are the cells of the grid. In a two-dimensional circular grid, each space between two adjacent circles is further subdivided, such as into quadrants. In other words, the grid may be formed by circles, the space between two circles being (in a one-dimensional circular grid) or including (in a two-dimensional circular grid) a cell of the grid. If a two-dimensional circular grid is used, each space between two circles can be divided into quadrants, for example, so that there are multiple cells in the space between the two circles. Each circle of the circular grid can represent a distance. For example, an active transceiver in a cell placed between the center point of the circular grid (i.e. where the mobile transceiver is located) and the first circle (from the center point) may have a distance of at most n meters from the mobile transceiver , the active transceiver placed in the cell between the first circle and the second circle may have a distance of more than n meters and at most 2 n meters from the active transceiver, and so on. This grid is used in the table of Figure 2.

In an embodiment, determining 110 more than one environment model may include obtaining information about one or more active transceivers. For example, the method may include collecting information about the location of the one or more active transceivers via wireless messages (e.g., wireless vehicle-to-vehicle messages, if the active transceiver is a vehicle) via the one or more active transceivers. For example, periodic status messages of one or more active transceivers may be processed to collect information about the location of the one or more active transceivers. Accordingly, the method may include receiving wireless transmissions from one or more active transceivers. The mobile transceiver may receive wireless transmissions from one or more active transceivers (eg, other vehicles/transceivers). Based on the received wireless transmissions, the location of one or more active transceivers can be determined. For example, the wireless transmission may include information about the location of the active transceiver that has sent the corresponding wireless message. The method may include determining an environmental model of the one or more active transceivers based on the received wireless transmissions of the one or more active transceivers. More generally, the method may include generating an environment model based on the determined locations of the one or more active transceivers.

In various embodiments, the method may include receiving at least a portion of at least a subset of at least a subset of the plurality of environment models from another mobile transceiver (e.g., from a mobile transceiver of one or more active transceivers). For example, after determining their environment models, mobile transceivers may share information with other mobile transceivers, such as by broadcasting the corresponding environment models. In other words, the method may include broadcasting the environment model to other mobile transceivers.

Based on the information collectively collected by the mobile transceivers, each mobile transceiver/vehicle can perform predictions of the future QoS of the wireless links it maintains with other mobile transceivers.

In an embodiment, multiple environmental models for one or more active transceivers are determined at multiple points in time. Each of the plurality of environment models may represent one or more active transceivers in an environment in which the transceiver is moved at one of the plurality of points in time. Multiple environment models can be generated over time (e.g. periodically). Figure 2 shows an example of multiple environment models generated at multiple points in time (eg, every 0.5 seconds). In the context of this disclosure, the term "at or for (future) one or more points in time" simply means that an action was performed in relation to the one or more points in time, and does not necessarily mean that the action is performed at exactly the same point or points in the future. However, in some cases, corresponding actions may be performed at the same point in time, such as where the quality of service is determined (rather than predicted) at a particular point in time.

The method includes determining 120 a predicted future environment model for one or more active transceivers at future points in time using 125 time series predictions for the plurality of environment models. Typically, time series forecasting predicts the development of one or more values over time intervals, including multiple points in time, based on historical data about one or more values. In other words, the trend of one or more values can be predicted based on historical data about the values, and the time series of developments of the values can be predicted. In an embodiment, one or more values may represent (i.e. form) an environment model. For example, each environment model can be represented by multiple values. For example, each environment may include numerical information about the number of active transceivers per cell of the grid, where the number of active transceivers per cell of the grid is a value for which time-series predictions are performed. Thus, the plurality of environmental models for one or more active transceivers in the environment of the mobile transceiver may be historical data (representing predicted future environmental models) for one or more values.

Different methods exist to perform time series forecasting. In some embodiments, statistical based methods may be used. For example, time series forecasting can be performed 125 based on a statistical fitting function or based on a temporal autocorrelation function. In a statistical fit function, you can predict the trend of a value by performing a fit to historical data about that value. Temporal autocorrelation is an autocorrelation function performed on a time series to predict the development of the time series at a future time point or period. Both functions can be applied to the above targets. For example, a statistical fit function or temporal autocorrelation function can be applied to multiple environmental models to determine a predicted future environmental model (at a future time point, or at least two future time points).

Alternatively, machine learning can be used to perform time series forecasting. In other words, time series forecasting 125 is performed using another machine learning model. Typically, as will be described later, the prediction of numerical values via a machine learning model can be performed using regression-based machine learning algorithms. To train another machine learning model, multiple environment models (represented by numerical values) can be divided into training input, training output, and unused environment models. For example, if a future environment model is to be predicted at points in time outside of a predetermined time interval, then for each training sample, a subset of multiple environment models can be selected as training inputs for training another machine learning model, and the Environment models determined for times other than a predetermined time interval (in the future) from a subset of the environment models are used as training outputs. This subdivision can be repeated in a sliding window fashion over multiple environment models. Based on this training, another machine learning model may be configured to generate a predicted future environment model for a future point in time that is a predetermined time interval that is input into the other machine learning model. Again, this other machine learning model can be used on a subset of the plurality of environment models in a sliding window fashion to determine predicted future environment models for at least two/more future points in time (or rather, at least two / Multiple predicted future environmental models). The same concept can be applied to statistics-based functions.

Both of these approaches attempt to predict the future development of the environmental model (i.e., up to a future point in time) so that the predicted future environmental model can then be used to predict future quality of service. In other words, the time series forecast can be determined 125 such that the progress of the environmental model towards the predicted future environmental model is predicted. In any case, the prediction of the future environment model may attempt to model the movement of one or more active transceivers and the corresponding movement of the mobile transceiver (and another mobile transceiver).

Counterintuitively, time series forecasting may not be directly applied to forecasts of future quality of service (as this may not be reliable without some additional input), but rather to an environmental model that underlies forecasts of future quality of service. In other words, time series forecasting can produce a predicted future environment model, where future service quality is predicted based on the predicted future environment model.

The method includes predicting 130 future quality of service of the wireless communication link at future points in time using a machine learning model. Predicted future environment models are used as input to machine learning models135. The output of the machine learning model may be the future quality of service of the wireless communication link, or the future quality of service of the wireless communication link may be predicted based on the output of the machine learning model.

Machine learning refers to algorithms and statistical models that computer systems can use to perform specific tasks without using explicit instructions, but relying on models and reasoning. For example, in machine learning, data transformations inferred from analysis of historical and/or training data may be used instead of rule-based data transformations. For example, the content of an image can be analyzed using a machine learning model or using a machine learning algorithm. In order for the machine learning model to analyze the content of the image, the machine learning model can be trained using the training image as input and the training content information as output. By training a machine learning model with a large number of training images and associated training content information, the machine learning model "learns" to identify the content of the images, and thus the machine learning model can be used to identify the content of images not included in the training images. The same principle can be used for other types of sensor data: by training a machine learning model with the training sensor data and the desired output, the machine learning model "learns" a transformation between the sensor data and the output that can be used to Learn the model's untrained sensor data to provide output. In an embodiment, the machine learning model is trained to provide a translation between the environment model and the quality of service of the wireless communication link predicted from the environment model. In other words, the machine learning model is trained to correlate the environment model with the predicted quality of service of the wireless communication link.

Use the training input data to train a machine learning model. The example specified above uses a training method called "supervised learning". In supervised learning, a machine learning model is trained using multiple training samples, where each sample may include multiple input data values and multiple expected output values, that is, each training sample is associated with an expected output value. By specifying both training samples and expected output values, the machine learning model "learns" which output value to provide based on input samples that are similar to the samples provided during training. Supervised learning can be based on supervised learning algorithms, such as classification algorithms, regression algorithms, or similarity learning algorithms. Classification algorithms can be used when the output is restricted to a finite set of values, i.e. when the input is classified as one of the finite set of values. Regression algorithms can be used when the output may have any value (within a certain range). Similarity learning algorithms are similar to both classification and regression algorithms, but are based on learning from examples using a similarity function that measures the similarity or relatedness of two objects.

Typically, machine learning models are trained to provide information about the predicted quality of service for a given environment model. For example, the predicted quality of service of a wireless communication link can be represented by a numerical value, which can be obtained using a regression-based machine learning algorithm. For example, the predicted quality of service may involve at least one of inter-packet reception time, packet error rate, delay, and data rate, all of which may be represented by numerical values. Accordingly, a machine learning model can be trained to implement (i.e., be based on) a regression algorithm, for example, to predict at least one of inter-packet reception time, packet error rate, latency, and data rate of a wireless communication link.

In some embodiments, the machine learning model is trained specifically for mobile transceivers. Alternatively, the machine learning model may be a general machine learning model applicable to different mobile transceivers. In either case, the method may include training a machine learning model, or the machine learning model may be trained by another entity (for which another method may be used, as shown in Figure Id).

Generally, a machine learning model can be trained to associate (predicted) quality of service with an environment model of one or more active transceivers in the environment of the mobile transceiver. Therefore, the environment model and the corresponding quality of service can be used to train the machine learning model. Typically, a machine learning model is trained using multiple training samples. In an embodiment, multiple environmental models for one or more active transceivers are determined at multiple points in time. Accordingly, the method may include determining 140 the quality of service of the wireless communication link at multiple points in time. In other words, for multiple points in time, the quality of service (from one or more active transceivers in the environment of the mobile transceiver) can be determined. For example, quality of service may be determined by determining metrics of the wireless communication link, such as the inter-packet reception time, packet error rate, latency, and/or data rate of the wireless communication link. The method may include training 145 the machine learning model using the plurality of environmental models at the plurality of time points as training inputs and the quality of service of the wireless communication link at the corresponding plurality of time points as training outputs of the training of the machine learning model. In other words, the environment models of the multiple environment models can be used together with the corresponding quality of service as training samples for the training of the machine learning model.

In some embodiments, the output of the machine learning model may be something other than the "raw" values of various quality of service characteristics. For example, some important characteristics such as the packet-to-packet reception time (PIR time) of the wireless link are usually at their minimum value (i.e. the transmission time) because the corresponding transmission is successful the first time. Training a machine learning model in this case produces a machine learning model that is biased towards this minimum. This can be avoided by training a machine learning model to output proxy values that can be used to determine actual quality of service characteristics. One such proxy is a probability distribution, which models how probable a given value of a quality of service characteristic (such as inter-packet reception time, packet error rate, delay, and data rate) is probable for a given environmental model. In other words, a machine learning model can be trained to provide a probability distribution of predicted quality of service given an environmental model. For example, as shown in conjunction with Figures 3a to 3d, some characteristics of quality of service, such as inter-packet reception time (PIR time), may be modeled using an exponential distribution of probabilities. This exponential distribution can be modeled using the following formula, where γ is the quality of service characteristic and λ is the rate.

This rate can be modeled as a function of the environment model (ie, the distribution of one or more active transceivers in the environment of the mobile transceiver) and the distance between the transmitter and receiver (inter-antenna distance, IAD) λ( x).

The rate is indicative of an (exponential) probability distribution of a characteristic of the quality of service (eg PIR time). Thus, a machine learning model can be trained to provide (predicted) rates of (predicted) quality of service for a given environment model, and thus probability distributions. The method may include determining a predicted future quality of service for the wireless link based on the rate/probability distribution provided by the machine learning model. Additional details can be found in the examples provided in conjunction with Figures 3a to 3d, where a suitable machine learning model is trained based on simulated data.

This rate can then be used to determine predicted future quality of service. For exponential distributions, 1/λ can be used to obtain predicted future quality of service values. The quantile function of the exponential distribution (also known as the inverse cumulative distribution function) can be used to obtain the quantile of the predicted future quality of service value:

For example, use p=0.75 to obtain the third quantile, and use p=0.9 to obtain the ninth quantile.

In various embodiments, future quality of service is predicted for multiple points in time in the future (i.e., at least two points in time in the future). For example, the method may include predicting future developments in quality of service at multiple points in time in the future. By determining a predicted future environment model (or rather, a plurality of predicted future environment models) at multiple points in the future, and using the predicted future environment model as an input to a machine learning model, many future Embodiments of the present disclosure are scaled at each point in time. In other words, by determining 120 a predicted future environment model of the one or more active transceivers at at least two future points in time, and using 135 the predicted one or more active transceivers at at least two future points in time The future environment model is used as an input to the machine learning model, which can predict the future quality of service of the wireless communication link for at least two future time points.

In the following, another method is shown that can be used to train a machine learning model independently of the method provided above. Figure Id shows a flowchart of an embodiment of a method for training a machine learning model. The method of FIG. Id includes determining 140 the quality of service of the wireless communication link at the plurality of time points; and using the plurality of environment models at the plurality of time points as training input and the wireless communication link at the corresponding plurality of time points The quality of service is used as the training output of the training of the machine learning model to train 145 machine learning models.

Machine learning algorithms are usually based on machine learning models. In other words, the term "machine learning algorithm" can refer to a set of instructions that can be used to create, train, or use a machine learning model. The term "machine learning model" may refer to a data structure and/or set of rules representing, for example, learned knowledge based on training performed by a machine learning algorithm. In an embodiment, the use of a machine learning algorithm may imply the use of an underlying machine learning model (or multiple underlying machine learning models). The use of a machine learning model may mean that the machine learning model and/or the data structure/rule set that is the machine learning model is trained by a machine learning algorithm.

For example, the machine learning model may be an artificial neural network (ANN). ANNs are systems inspired by biological neural networks such as those found in the brain. An ANN consists of multiple interconnected nodes and multiple connections (so-called edges) between nodes. There are generally three types of nodes, namely input nodes that receive input values, hidden nodes that are (only) connected to other nodes, and output nodes that provide output values. Each node can represent an artificial neuron. Each edge can transmit information from one node to another. The output of a node can be defined as a (non-linear) function of the sum of its inputs. A node's input can be used in edge-based or "weights" functions of the node that provides the input. The weights of nodes and/or edges can be adjusted during the learning process. In other words, training of the artificial neural network may include adjusting the weights of the nodes and/or edges of the artificial neural network, i.e. to achieve a desired output for a given input. In at least some embodiments, the machine learning model may be a deep neural network, such as a neural network comprising one or more layers of hidden nodes (i.e., hidden layers), preferably multiple layers of hidden nodes.

该架构还包括输出层，该输出层输出率λ和因此服务质量的概率分布。Figure 3c shows a schematic diagram of an exemplary architecture of a machine learning model. The machine learning model consists of an input layer with input nodes for the distance d between a mobile transceiver and another mobile transceiver, and for active transceivers in cells of a circular grid of width g (by A _1,g ... An _,g represents the number of), where n is the number of cells of the circular grid under consideration. The architecture also includes n _h hidden layers, each with n _n nodes

The architecture also includes an output layer that outputs a probability distribution of rate λ and thus quality of service.

In some embodiments, the machine learning model may be a support vector machine. A support vector machine (i.e., a support vector network) is a supervised learning model with an associated learning algorithm that can be used to analyze data, for example, in classification or regression analysis. A support vector machine can be trained by providing an input with multiple training input values belonging to one of two categories. A support vector machine can be trained to assign new input values to one of two classes. Alternatively, the machine learning model may be a Bayesian network, which is a probabilistic oriented acyclic graphical model. Bayesian networks can use directed acyclic graphs to represent a set of random variables and their conditional dependencies. Alternatively, the machine learning model may be based on genetic algorithms, which are search algorithms and heuristic techniques that simulate the process of natural selection.

Device 10 and mobile transceivers 100, 102, 104 (e.g., vehicles or entities) may communicate through a mobile communication system. The mobile communication system may for example correspond to one of the 3rd Generation Partnership Project (3GPP) standardized mobile communication networks, wherein the term mobile communication system is used as a synonym for mobile communication network. Thus, messages (input data, measurement data, control information) can be communicated through multiple network nodes (e.g., Internet, routers, switches, etc.) and the mobile communication system, which creates the delays or delays considered in the embodiments.

The mobile or wireless communication system may correspond to a fifth-generation mobile communication system (5G or new radio), and may use millimeter wave technology. A mobile communication system may correspond to or include, for example, Long Term Evolution (LTE), Long Term Evolution-Advanced (LTE-A), High Speed Packet Access (HSPA), Universal Mobile Telecommunications System (UMTS) or UMTS Terrestrial Radio Access Network (UTRAN), Evolved UTRAN (e-UTRAN), Global System for Mobile Communications (GSM) or Enhanced Data Rates for GSM Evolution (EDGE) networks, GSM/EDGE Radio Access Network (GERAN) or networks with different standards (eg Global Interoperability for Microwave Access) A mobile communication network operating (WIMAX) network IEEE 802.16 or Wireless Local Area Network (WLAN) IEEE 802.11), typically an Orthogonal Frequency Division Multiple Access (OFDMA) network, Time Division Multiple Access (TDMA) network, Code Division Multiple Access (CDMA) network , Wideband CDMA (WCDMA) network, Frequency Division Multiple Access (FDMA) network, Spatial Division Multiple Access (SDMA) network, etc.

The service provisioning may be performed by a network element such as a base transceiver station, a relay station or a UE that coordinates the service provisioning eg in a cluster or group of multiple UEs/vehicles. The base station transceiver is operable or configured to communicate with one or more active mobile transceivers/vehicles, and the base station transceiver may be located at another base station transceiver (eg, a macrocell base station transceiver or a small cell base station transceiver) in or near the coverage area. Accordingly, embodiments may provide a mobile communication system comprising two or more mobile transceivers/vehicles 100, 102 and one or more base station transceivers, wherein the base station transceivers may establish macro cells or small cells, such as pico cells, Metro cells or femto cells. A mobile transceiver or UE may correspond to a smartphone, cellular phone, laptop, notebook, personal computer, personal digital assistant (PDA), universal serial bus (USB) stick, car, vehicle, road participant, traffic entities, transport infrastructure, etc. According to 3GPP terminology, a mobile transceiver may also be called a user equipment (UE) or a mobile phone. For example, the mobile transceiver, another mobile transceiver, and/or at least a subset of the one or more active transceivers may be a vehicle, such as a land vehicle, on-highway vehicle, sedan, automobile, off-highway vehicle, motor vehicle, truck, or truck.

Base transceiver stations may be located in fixed or stationary parts of a network or system. A base transceiver station may be or correspond to a remote radio head, transmission point, access point, macro cell, small cell, micro cell, femto cell, metro cell, or the like. The base station transceiver may be a wireless interface to a wired network that enables radio signal transmission to the UE or mobile transceiver. Such radio signals may conform, for example, to radio signals standardized by 3GPP, or generally conform to one or more of the systems listed above. Thus, a base transceiver station may correspond to a NodeB, eNodeB, gNodeB, base transceiver station (BTS), access point, remote radio head, relay station, transmission point, etc., which may be further subdivided into remote units and central units.

A mobile transceiver or vehicle 100 may be associated with a base transceiver station or cell. The term cell refers to the coverage area of radio services provided by base transceiver stations such as NodeBs (NBs), eNodeBs (eNBs), gNodeBs, remote radio heads, transmission points, and the like. A base transceiver station may operate one or more cells on one or more frequency layers, and in some embodiments, a cell may correspond to a sector. For example, sectors may be implemented using sector antennas that provide the characteristic of covering angular segments around a remote unit or base transceiver station. In some embodiments, the base transceiver station may, for example, operate three or six cells covering a sector of 120° (in the case of three cells), 60° (in the case of six cells), respectively. The base transceiver station can operate multiple sectorized antennas. In the following, a cell may refer to the respective base transceiver station that generated the cell, or similarly, a base transceiver station may refer to a cell generated by the base transceiver station.

In an embodiment, the apparatus 10 may be included in a server, a base station, a NodeB, a UE, a mobile transceiver, a relay station, or any service coordination network entity. It should be noted that the term network element may include multiple sub-elements, such as base stations, servers, and the like.

In an embodiment, the one or more interfaces 12 may correspond to any means for obtaining, receiving, transmitting or providing analog or digital signals or information, such as any connectors, contacts, pins, registers, input ports, outputs Ports, conductors, channels, etc. that allow signals or information to be provided or obtained. The interface may be wireless or wired, and it may be configured to communicate with further internal or external components, i.e. to send or receive signals, information. The one or more interfaces 12 may include additional components to enable corresponding communications in the mobile communication system, which may include transceiver (transmitter and/or receiver) components, such as one or more low noise amplifiers (LNAs) , one or more power amplifiers (PA), one or more duplexers, one or more duplexers, one or more filters or filter circuits, one or more converters, one or more mixers frequency converter, correspondingly adapted radio frequency components, etc. One or more interfaces 12 may be coupled to one or more antennas, which may correspond to any transmit and/or receive antennas, such as horn antennas, dipole antennas, patch antennas, sector antennas, and the like. The antennas may be arranged in defined geometric arrangements, such as uniform arrays, linear arrays, circular arrays, triangular arrays, uniform field antennas, field arrays, combinations thereof, and the like. In some examples, one or more of the interfaces 12 may be used for the purpose of sending or receiving or sending and receiving information, such as information, input data, control information, additional information messages, and the like.

A respective one or more interfaces 12 are coupled to respective control modules 14 at the device 10, as shown in Figure 1c. In embodiments, control module 14 may use one or more processing units, one or more processing devices, any means for processing such as a processor, computer, or programmable hardware operable with correspondingly adapted software components) to achieve. In other words, the described functions of control module 14 may also be implemented in software, which is then executed on one or more programmable hardware components. Such hardware components may include general purpose processors, digital signal processors (DSPs), microcontrollers, and the like.

In embodiments, communication (i.e. sending, receiving, or both) may occur directly between the mobile transceivers/vehicles 100, 102, such as forwarding incoming data or control information to/from a control center. Such communication may utilize a mobile communication system. Such communication can take place directly, for example by means of device-to-device (D2D) communication. Such communication may be carried out using the specifications of the mobile communication system. An example of D2D is direct communication between vehicles, also known as vehicle-to-vehicle communication (V2V), vehicle-to-vehicle communication, dedicated short-range communication (DSRC), respectively. Technologies that support this D2D communication include 802.11p, 3GPP systems (4G, 5G, NR and above), and more.

In an embodiment, the one or more interfaces 12 may be configured to communicate wirelessly in a mobile communication system. For this purpose, radio resources are used, such as frequency, time, code and/or space resources, which can be used for wireless communication with the base transceiver station as well as for direct communication. The allocation of radio resources can be controlled by the base transceiver station, i.e. determining which resources are used for D2D and which are not used. Here and below, the radio resources of the respective components may correspond to any conceivable radio resources on the radio carrier, and they may use the same or different granularity on the respective carriers. Radio resources may correspond to resource blocks (such as RBs in LTE/LTE-A/LTE-unlicensed (LTE-U)), one or more carriers, sub-carriers, one or more radio frames, radio frames, radio Time slots, one or more code sequences potentially with corresponding spreading factors, one or more spatial resources (such as spatial subchannels, spatial precoding vectors), any combination thereof, and the like. For example, in direct cellular vehicle-to-anything (C-V2X), where V2X includes at least V2V, V2-Infrastructure (V2I), etc., transmissions according to 3GPP Release 14 and above can be managed by the infrastructure (so-called Mode 3) Or run in UE.

Further details and aspects of the method, apparatus or mobile transceiver are referred to in connection with the concepts presented or one or more examples described above or below (e.g. Figures 2 to 3d). The method, apparatus or mobile transceiver may include one or more additional optional features corresponding to one or more aspects of the presented concepts or one or more examples described above or below.

Various embodiments relate to a method of increasing the prediction time domain in QoS prediction using time series prediction. In the context of collaborative driving, when quality of service (QoS) conditions change, prediction of future QoS enables vehicle applications. In fact, when no predictive QoS (PQoS) is provided, the application can only react to changes and is therefore limited to the lower bound performance of the communication system. The PQoS system may operate on the vehicle (communication node) using a radio access technology (RAT) such as LTE-V or 5G-V2X or IEEE 802.11p in its standalone mode. Combinations of these techniques can also be applied to multi-RAT systems. In such a PQoS system, vehicles can exchange information about the communication surroundings in order to provide PQoS. Accordingly, embodiments may relate to methods for predicting QoS in the future.

In other methods such as the paper "Prediction of Packet Inter Reception Time for Platooning using Conditional ExponentialDistribution" by Jornod, El Assad, Kwoczek and Kürner, a statistical relationship between ambient density and inter-packet reception time is provided. However, in the method shown in this paper, the prediction is only instantaneous, and the embodiments aim to predict the future quality of service. More specifically, in a statistical sense, forecasting involves the modeling of instantaneous data and the forecast prediction of future values. According to this definition, forecasts are made using time series forecasts, and the results of forecasts are used to predict future quality of service.

Embodiments may provide a method for forecasting QoS in the future using historical data and time series forecasting.

The method includes one or more of the following features:

1. Collect transmission data (i.e. wireless transmissions) from receivers (i.e. mobile transceivers) and the locations of surrounding communication nodes (i.e. one or more active transceivers).

2. Calculate the QoS metrics of interest from this receiver (ie Quality of Service)

3. Model the surrounding communication environment (e.g. by generating an environment model), e.g. to obtain indicators of the surrounding communication environment.

4. Matching the calculated QoS metrics of the transmitted data to a model of the surrounding communication environment in terms of the time and location of the transmitter and receiver (e.g., linking the quality of service to the corresponding environment model), e.g., to enable estimation The correlation between the indicators of the surrounding environment and the indicators of QoS.

5. Train a distribution model (e.g. a machine learning model) on the collected data (see table)

6. Perform time-series forecasting in the future to predict future QoS in the forecasting time domain (e.g., use time-series forecasting to determine a predicted future environment model), e.g., be able to base on the correlation between the indicators of the surrounding environment and the indicators of QoS change in nature to predict QoS

7. Change the prediction time domain to obtain a dynamic link QoS map along time/direction (e.g. to determine a predicted future environment model at least two time points in the future), e.g. to be able to create a QoS map that enables Make decisions about adjusting app settings.

8. Repeat one or more previous features with accumulated data to improve learning (e.g., reinforcement learning)

Communication environment data collection can be performed by using a radio receiver as a sensor or by sharing this information with the communication. Modeling of the surrounding communication environment may be performed using a grid-like abstraction approach to radio activity. For example, the surrounding communication environment may be modeled using a grid-like division of the environment, with radio activity numerical levels assigned according to the number of communicating vehicles and the periodicity of their messages. For example, future time series forecasting may be performed to predict future QoS in the forecast time domain by using changes in correlation between previous environmental predictor values and previous QoS values. Time series forecasting can be performed by using samples from the past t ₀ -t _p to t ₀ to simulate ( FIG. 2 210 ) the distribution of QoS at t ₀ +t _f ( FIG. 2 230 ) (where t ₀ is the current time (Fig. 2 220), t _p is the lookback time, eg 10s, and t _f is the prediction time domain). Changes in the prediction time domain can result in multi-link QoS maps, one for each future time step. It may change with time/displacement, in that sense it is dynamic. The calculated and predicted QoS indicators can be packet error rate (PER), inter-packet reception (PIR) time, delay, and data rate. For example, time series forecasting may use statistical forecasting methods, such as multilayer perceptrons (eg, as shown in Figure 3c).

Figure 2 provides an example of data used to train a predictive model, e.g. according to one of the examples provided in conjunction with Figures 3a to 3d. Figure 2 shows a table of the development of quality of service attributes related to the environment model over time. The table includes various columns, the first column shows the timestamp (time), the second and third columns show the identities of the destination and source nodes (dst and src), the fourth column shows the size of the packet (size), the fifth column The column shows the PIR value (pir), the sixth column shows the distance between the transmitter and receiver (dist, which remains fairly constant), and the seventh and subsequent columns show the geographic area around the receiver (src) ( That is, the number of nodes d* in a cell or torus of a circular grid, where * represents the maximum distance (eg, d30 = 0 up to 30 meters, d60 = greater than 30 up to 60, etc.). Rectangle 210 defines the lookback data, rectangle 220 defines the current time step (defining a predefined time interval to the prediction time domain), and rectangle 230 defines the target prediction time domain, e.g., 8 seconds. This "sliding window" can be applied to all time steps in order to train a predictive model, which effectively becomes a predictive model when we look at future data.

In various embodiments, the method includes monitoring an indicator of the surrounding communication environment, such as the number of surrounding communicating vehicles in a cell of the grid, and correlating the change with a change in QoS. For example, the method may include predicting QoS (suitable for V2V communication) based on a correlation between changes in indicators of surrounding communication vehicles and changes in QoS. The method may include using information about the surrounding communication environment for the V2V link. The method may include creating a QoS map. The method may include performing a match between communication environment metrics and QoS changes. The method may include predicting a future value of QoS using past changes in the correlation between environmental predictors and QoS. Embodiments may use aggregated communication activity at a higher layer (message level) where we wish to predict the impact of hidden nodes and channel overload due to surrounding communicating vehicles.

Figures 3a to 3d show schematic diagrams related to the training of machine learning models. In the following, two similar approaches are presented, both involving the training of a machine learning model adapted to provide a probability distribution on the predicted quality of service for a given environment model. While machine learning models are trained using simulated environment models, the same principles can be applied to data collected in "real" mobile transceivers/vehicles.

和它们的指数

来限定。在G.Jornod、A.El Assaad、A.Kwoczek和T.Kürner在2019年第28届IEEE欧洲网络和 通信会议中的“Packet inter-reception time modelingfor high-density platooning in varying surrounding traffic density”中给出了这些特征的进 一步描述。在模拟中，计算了在PIR、400m半径内的车辆的总数量和 IAD之间的关系的表示。在这种情况下，当周围通信车辆的数量超过 150辆时，PIR急剧增加。总数对目标也有影响；当IAD变得大于100m 时，这种影响急剧降低。In the following, the background of these two methods is presented. An interesting application of AQoSA (Adaptive Quality of Service Adaptation) is High Density Platooning (HDPL): this is a cooperative vehicle application in which vehicles coordinate their control to achieve a small inter-vehicle distance of less than 15m ( IVD). Within the scope of AQoSA, prediction systems that can predict future QoS are required. The predictive system can use its knowledge of its surroundings to function. The following scenario features five-truck HDPL driving at 25 m/s on an oval multi-lane test track. This cohort targets IVDs between 5 and 25m. It is coordinated by broadcasting a platooning control message (PCM), which is a 700B message transmitted at a rate of 20Hz using IEEE 802.11p Radio Access Technology (RAT). To challenge the communication system, increasing densities of vehicular traffic in the opposite lanes are introduced. Within a range of 400m around the transmitter, this surrounding traffic reaches more than 200 vehicles. Surrounding vehicles utilize the same RAT to broadcast 400B CAMs (co-aware messages) at a rate of 10Hz. The mobility of surrounding traffic is modeled in Simulation of Urban Mobility (SUMO), a simulation tool that takes full advantage of its ability to generate random vehicles with a specific density. This traffic simulator is combined with the network simulator ns-3, which utilizes its IEEE 802.11p model to operate wireless transmissions. This setup allows simulation of real vehicle motion as well as a fairly accurate model of the communication system. Using the tracking capabilities of the ns-3, a total of 106 transmission observations were collected. These observations include observed PIR times and environmental predictors described below. The PIR time, denoted γ, is defined as the duration between two consecutive messages in a pair of communication partners measured from the receiver. The PIR value is the result of consecutive transmission failures, which produces a low representation of a high PIR value. However, these higher values are very important for the application as they may limit the performance of the application. A direct prediction of γ most likely results in a systematic prediction for a PCM transmission period of 0.05 sec, as it is the smallest PIR and the most representative value. Instead, a PIR probability distribution can be modeled. The first set of features is the IAD d, the distance between the transmitter and receiver, calculated between their antennas. The second group includes or consists of the number of communicating vehicles _Dn,g =card(an _,g ) in the torus around the transmitter. The torus an _{, g} is determined by the difference of their radii

and their indices

to limit. In "Packet inter-reception time modeling for high-density platooning in varying surrounding traffic density" by G. Jornod, A. El Assaad, A. Kwoczek and T. Kürner at the 28th IEEE European Conference on Networking and Communications 2019 A further description of these features is provided. In the simulations, a representation of the relationship between the PIR, the total number of vehicles within a 400m radius and the IAD was calculated. In this case, when the number of surrounding communication vehicles exceeds 150, the PIR increases sharply. The total also has an effect on the target; this effect decreases sharply when the IAD becomes larger than 100m.

In the following, based on the learning of nonlinear functions providing exponentially distributed parameters, a predictive model is provided (which can implement the machine learning model introduced in connection with Figs. 1a to 1d).

The distribution of PIR can be regarded as a conditional exponential distribution. In the literature, PIR is shown to be well modeled by exponential distribution, lognormal distribution, Weibull distribution, gamma distribution, the last two distributions being preferred, which pass the 90% significance test. Therefore, in the following, PIR can be modeled with an exponential distribution. The exponential distribution can be seen as a special case of the gamma distribution, with the advantage of having a single parameter, the rate λ>0. Its probability density function (PDF) is expressed as:

And several rates of interest λ are shown in Fig. 3a, and their cumulative density function (CDF) in Fig. 3b. λ is parameterized by the spatial predictor:

where x is a quadratic polynomial combination of predictors d and _Dn,g with only interactions (eg, omitting ^d2 ). This modeling allows the calculation of PIR densities for any combination of predictors within the range of training data. The purpose of modeling is to approximate the nonlinear function λ(x) using the observed data. For this, an MLP regressor was used (see Fig. 3c). A Keras-based implementation was used, where the model was trained using the Adam optimization algorithm with negative log-likelihood as the loss function:

where Ω is the set of indices containing the training data. A k-fold cross-validation was performed for the choice of size and the number of hidden layers (n ∈ {1, 2, 3} and m ∈ {100, 500, 1000}, respectively) was applied using scikit-learn, k=8. A limitation of this model is that it does not take advantage of previous values of the predictors. In fact, the PIR time represents the time between two successful events: the source of consecutive grouping failures may occur between two observations. However, considering that the predictors are moving vehicles, their motion is smooth enough that the model is robust to this limitation. The combined use of these two python libraries enables rapid prototyping of learning strategies. The MLP regressor was chosen for its ability to capture nonlinearities highlighted during the data exploration phase.

Another method is proposed below. In the following, the prediction of the PIR distribution using the spatial distribution of communication nodes is shown. Previous studies have shown that PIR can be modeled with an exponential distribution. In addition, the IAD and surrounding communication traffic density changes the performance of the communication system. In the following concepts, these two premises are used to build a model that relates the distribution of predicted targets to changes in the spatial distribution of nodes. Subsequently, the fast learning of this relationship in a full-scale system-level Vehicle Ad hoc Network (VANET) simulation featuring a five-truck HDPL is investigated.

Using the tracking capabilities of the ns-3, a total of 4·10 ⁷ transmission observations were collected. Of these, 2.5·10 ⁶ has queue members on the receiving side, and 9·10 ⁵ describes intra-queue communication. These observations include the observed PIR time, source and destination, and information about the location of surrounding vehicles. Figure 3d shows an exemplary distribution of PIR. PIR time shows a heavy-tailed distribution, which motivates further modeling of its distribution, rather than simply predicting values. In effect, the PIR value is the result of successive transmission failures, which produce a low representation of a high PIR value. However, these higher values are very important for the application as they may limit the performance of the application. As in Jornod, El Assaad, Kwoczek and Kürner, "Prediction of Packet Inter-Reception Time for Platooning using ConditionalExponential Distribution," in 16th Int. IEEE Symp. Wireless Commun. Sys. (ISWCS), 2019, pp. 265–270 Emphasized that simple predictions most likely resulted in the most representative value of 0.05 sec for systematic predictions. However, the distribution of the number of vehicles differs from previous studies. In effect, vehicles are introduced into the scene in a random fashion. Vehicles then follow random routes in the scene, possibly clustering, an experimental design choice to investigate the robustness of the method to unknown or less frequent situations.

In previous studies, the importance of considering the interaction between the IAD and the number of surrounding vehicles was shown. The average PIR for the combination of IAD and car number interval was simulated. A sharp increase in the average PIR value above 125m was observed. In general, it also increases with both the number of surrounding vehicles and the IAD. In the region between 80 and 100 vehicles, higher PIR values were observed. This region and the second gradient it induces were not seen in previous results. Indeed, in this previous study, surrounding vehicles were restricted to straight roads. In this new scene, the spatial distribution of surrounding vehicles changes more dramatically. This new pattern and the appearance of a second incremental path from the origin to the top of the ellipse are the result of this new distribution. This further motivated the modeling of this distribution and the use of nonlinear regression models.

In the following, modeling strategies are introduced. The first step is to calculate the target key performance indicator (KPI), PIR time (i.e. quality of service). The second step is the modeling of the surrounding communication environment (ie, the environment model), which includes location log preprocessing and abstraction of the environment. The third step is to formalize the relationship between these environmental features and the distribution of our objective (i.e. the generation of training data). The fourth and final step is to create a strategy for the learning of this relationship. In this example, the prediction target is PIR time. As mentioned earlier, this metric has received increasing attention in VANET research. Within the scope of HDPL, it measures the time when a vehicle cannot rely on communications to coordinate with other vehicles. Furthermore, modern control systems are able to cope with low update input rates due to predictive algorithms. However, they perform poorly on irregular inputs. This regularity is captured by the PIR distribution. PIR is the time difference between the reception of two messages measured from the receiver of each source. This KPI is studied for cohort members for all received information. Because these messages are periodic, the PIR time is a multiple of the transmission period, plus or minus the delay experienced. It reflects the number of consecutively discarded messages.

中的车辆的数 量，该车辆是以接收器为中心的具有半径差

的同心环面中的通 信车辆，

是环面的指数。In the following, environmental characteristics are introduced. One goal is to consider the location of all communicating vehicles (ie one or more active transceivers). The main challenge is that the number of inputs is variable (theoretically, this number can span from a few cars to infinity if the focus is not within a certain range). Furthermore, most forecasting methods require fixed-size and ordered inputs even when the scope is narrowed down to a specific range. As a result, an appropriate aggregation method (to represent the environment model) for surrounding communication environment information is used. In G. Jornod, A. El Assaad, A. Kwoczek and T. Kürner, "Packet inter-reception time modeling for high-density platooning in 28th IEEE European Conference on Networks and Communications (EuCNC) 2019, pp. 187–192 In "varying surrounding traffic density", a torus-based environment model (ie, based on a circular mesh) is introduced, such as shown in Figures 3a to 3d. The core idea is to divide the space in concentric circles around the receiver and count the number of vehicles present in the formed torus. The radius of these circles is a multiple of the particle size parameter r. As a result, _An,r is contained in

The number of vehicles in which the vehicle is centered on the receiver with a radius difference

communication vehicles in the concentric torus,

is the index of the torus.

Using this model, the effect of the spatial distribution of surrounding vehicles on the channel loading is captured. The torus is defined as:

_An,r =card(an _,r )

是周围车辆位置向量的集合，n是环面指数，r是其粒度值， 并且x_r是接收器的位置向量。环面划分设计用于高速公路使用情况和 在相对车道上车辆进入的特定场景。in

is the set of surrounding vehicle position vectors, n is the torus index, r is its granularity value, and x _r is the receiver's position vector. Torus partitioning is designed for highway usage and specific scenarios of vehicle entry in opposite lanes.

In this example, this modeling (ie, the environment model) is complemented by the introduction of sectorization. This division captures the location of interfering nodes, especially relative to the transmitter location, which may not be detected by the transmitter when the IAD is relatively high. The space is divided into _ns regular sectors of angle α centered on the receiver (ie, the center point of the circular grid) and aligned with the receiver-transmitter segment. The division method can be defined as:

是提供正x轴和向量x之间的角度的函数，β是产生Rx–Tx 和Rx–干扰车辆向量之间的角度的函数。

是角度α＝2π/n_s的第m 个扇区

内的车辆的数量。最后，引入偏移π/2，以便表示当n_s＝2时 的前/后划分，而不是左/右划分。通过结合基于环面的模型和基于扇 形的模型，获得了所谓的环面-扇区模型。其部段由环面和扇区的交点 限定。在该示例中，

表示环面

与扇区

的交点。类似地，包 含在部段中的车辆组及其基数被定义为：in,

is the function that provides the angle between the positive x-axis and the vector x, and β is the function that produces the angle between the Rx–Tx and Rx–disturbing vehicle vectors.

is the mth sector of the angle α=2π/ _ns

the number of vehicles inside. Finally, an offset π/2 is introduced in order to represent the front/back division when _ns = 2, rather than the left/right division. By combining a torus-based model and a sector-based model, a so-called torus-sector model is obtained. Its segments are defined by the intersection of the torus and the sector. In this example,

represents a torus

with sector

intersection. Similarly, the groups of vehicles included in a segment and their cardinality are defined as:

和

在该示例中，扇区总是相 对于发射器定向，并且环面反映了干扰源的距离。当通过特征的多项 式组合的步骤与IAD组合时，该环境模型提供了解决隐藏节点问题的 可能性。这个问题可通过加权过程在目标和特征之间的关系的建模中 解决。It can be noticed that,

and

In this example, the sector is always oriented relative to the transmitter, and the torus reflects the distance to the interferer. When combined with IAD through the step of polynomial combination of features, this environment model offers the possibility to solve the hidden node problem. This problem can be addressed in the modeling of the relationship between objects and features through a weighting process.

表示周围的节点位置。这被区分为两种情况：全局知识和 局部知识。在全局知识的情况下，它包含模拟中的所有节点，并且表 示为

该集合可从传输日志和位置日志获得。在局部知识的情况下， 该集合被表示为

并收集了接收器在最后T秒内从其接收到CAM的 节点。根据信道负载，在具有较低T值的情况下，该集合中包含的节 点的数量可急剧减少。当没有实现集体感知系统时，该缩减的集合也 反映了周围通信环境的现实知识。同样，可使用传输日志来计算该集 合。在示例的评估中，T被设置为10秒。gather

Indicates the surrounding node location. This is distinguished into two cases: global knowledge and local knowledge. In the case of global knowledge, it contains all nodes in the simulation and is represented as

This collection is available from the transport log and the location log. In the case of local knowledge, the set is represented as

and collected the nodes from which the receiver received the CAM in the last T seconds. Depending on the channel load, with a lower value of T, the number of nodes contained in the set can be drastically reduced. This reduced set also reflects real-world knowledge of the surrounding communication environment when no collective perception system is implemented. Again, the transport log can be used to calculate the set. In the evaluation of the example, T is set to 10 seconds.

In this example, the prediction target is PIR time. Hereinafter, Γ denotes PIR as a random variable. As shown in Figure 3d, its distribution is heavy-tailed. This characteristic of the distribution may preclude the use of classical regression methods aimed at directly predicting PIR. In fact, since lower values are more representative than larger values, a simple prediction will result in a systematic prediction of the more representative value, which happens to be the transmission period. Instead, the distribution of target values can be predicted, which adds a distribution modeling step before solving the learning method. In the literature, PIR is shown to be well modeled by exponential distribution, lognormal distribution, Weibull distribution, gamma distribution.

The probability density function (PDF) of the exponential distribution is given as:

It has the advantage of having a single parameter, the rate λ. Therefore, the modeling task may be to find a suitable λ for the collected data. Note that in the absence of relative delay, the PIR time is a discrete variable. In M.E.Renda, G.Resta, P.Santi, F.Martelli, and A.Franchini, "IEEE 802.11p VANets: Experimentalevaluation of packet inter-reception time," Comput. Commun., Vol. 75, pp. 26–38, In 2016, PIR was modeled with a geometric distribution. An exponential distribution is a continuous analog of this distribution. It makes it possible to avoid the step of eliminating relative delays in the collected data.

The previous subsection described the environmental characteristics. Before this, the premise for studying the effects of the number of surrounding vehicles and IAD and their interactions was shown. The results show that the number of vehicles and IAD together affect the average PIR of the CAM information. Based on the number of vehicles around the receiver, on the one hand, and on the IAD, on the other hand, the conditional CDF is calculated. The first observation is that all provided CDFs resemble those of an exponential distribution. The second observation is that the rate λ varies with the number of vehicles n _v and IADd . The rate is also different when combining the two torus-divided intervals ( _nv,1 and _nv,1 ), which shows the effect of the spatial distribution of the surrounding nodes. This facilitates the parameterization of λ with environmental features. It is recommended to use λ as a function of a polynomial combination of spatial distribution features and IAD:

where x is a combination of quadratic polynomials with only interactions (eg omitting d ² ). Therefore, the remaining steps are to approximate the function λ(x) such that Γ ~ Exp(λ(x)). This process is called conditional density estimation (CDE).

In the following, a learning method (eg, for training a machine learning model) is shown. The goal might be to provide a flexible method that enables fast learning of the PIR distribution. The nonlinear function λ(x) can be approximated by a multilayer perceptron (MLP). The Keras interface can be fully exploited and combined with scikit-learn for hyperparameter optimization. Then, the selection of the number of hidden layers n _h and the number of intra-layer nodes n _n is made automatically by a cross-validated grid search. The model (i.e., the machine learning model) is trained using the Adam optimization algorithm with the negative log-likelihood as the loss function:

where Ω ₁ is the set of training data. The MLP is fed a polynomial combination of features (ie, a polynomial combination based on an environment model) and outputs a rate λ (which can be used to determine the probability distribution of the quality of service). Figure 3c illustrates this process for a torus-based environment model. The learning model has three parameters: the learning rate LR, the number of hidden layers n _h and the number of neurons in each layer n _n . For example, it can be assumed that there are the same number of neurons in each layer. The performance of the torus-sector model can be compared according to the parameters r and _ns . This was done using an MLP regressor on data collected by cohort members during the first 30 minutes (ie, multiple environmental models and corresponding quality of service). A k-fold cross-validation strategy is used to split between training and test sets. The strategy is applied to each model evaluated and is reported as the average performance. The model selection step may also include three learning parameters (learning rate, size and number of layers). The best performing model is reported for each combination of r and n _s . This best performing model is then used for the total simulation duration. The MLP regressor is iteratively trained with data collected by members of the training cohort. During this training process, the weights of the regressor are updated at each iteration. The performance of the model is evaluated against test data, which includes all observations collected by the remaining trucks for the duration of the simulation. The log-likelihood given above is used as the loss function for the training of the MLP model.

The evaluation found that the best performing torus-sector model parameters were r=30, _ns = _8LR =0.0001, _nn =1000 and nh=9 for the global scope and r=60, _ns =9 for the local scope 4, LR=0.0001, n _n =500 and n _h =10. The following values were evaluated: r=n30m, where n∈{1,2,3,4,6,13}, _ns∈ {1,2,4,8}, LR={0.1, 0.01, 0.001, 0.0001} , n _n ={50, 100, 500, 1000} and n _h ={2, 3, . . . , 10}.

The selected model is trained online throughout the simulation using the results of previous model comparisons. The fitted model is updated each time the training node receives a transmission. The performance of the model was performed on two datasets: (i) data collected up to the reception time of the node of interest, which is a subset of the training set; (ii) and all data collected by other cohort members.

Evaluations show that even after convergence, the model can continue to learn to improve its robustness.

Further details and aspects of this concept are mentioned in connection with the proposed concept or one or more examples described above or below (e.g., Figures 1a-2). The concept may include one or more additional optional features corresponding to one or more aspects of the proposed concept or one or more examples described above or below.

As already mentioned, in the embodiments, the corresponding methods can be implemented as computer programs or codes, which can be executed on corresponding hardware. Thus, another embodiment is a computer program having program code for performing at least one of the above-described methods when the computer program is executed on a computer, processor or programmable hardware component. Another embodiment is a computer-readable storage medium storing instructions that, when executed by a computer, processor or programmable hardware component, cause the computer to implement one of the methods described herein.

Those skilled in the art will readily recognize that the steps of the various methods described above may be performed by a programmed computer, for example, to determine or calculate the location of a time slot. Here, some embodiments are also intended to cover program storage devices, such as digital data storage media, that are machine- or computer-readable and encode a machine-executable or computer-executable program of instructions that perform the operations described herein. some or all steps of the method. The program storage devices may be, for example, digital memories, magnetic storage media such as magnetic disks and tapes, hard drives, or optically readable digital data storage media. The embodiments are also intended to cover computers programmed to perform the steps of the methods described herein or (field) programmable logic arrays ((F)PLAs) or (field) programmable logic arrays ((F)PLAs) programmed to perform the steps of the methods described above. Program Gate Array ((F)PGA).

The description and drawings merely illustrate the principles of the invention. Accordingly, it should be understood that those skilled in the art will be able to devise various arrangements that, although not explicitly described or shown herein, embody the principles of the invention and are included within its spirit and scope. Furthermore, all examples recited herein are primarily intended to be used expressly for teaching purposes only to assist the reader in understanding the principles of the invention and the concepts contributed by the inventors to advance the art, and are to be construed as not limited to these specifically recited examples and conditions . Furthermore, all statements herein reciting principles, aspects, and embodiments of the invention, as well as specific examples thereof, are intended to encompass equivalents thereof.

When provided by a processor, the functions may be provided by a single dedicated processor, a single shared processor, or multiple separate processors, some of which may be shared. Furthermore, explicit use of the terms "processor" or "controller" should not be construed to refer exclusively to hardware capable of executing software, and may implicitly include, but is not limited to, digital signal processor (DSP) hardware, network processors, Application Specific Integrated Circuits (ASICs), Field Programmable Gate Arrays (FPGAs), Read Only Memory (ROM) for storing software, Random Access Memory (RAM), and non-volatile storage. Other conventional or custom hardware may also be included. Their function may be implemented through the operation of program logic, through dedicated logic, through the interaction of program control and dedicated logic, or even manually, the specific technique being selectable by the implementer as more specifically understood from the context.

It should be understood by those skilled in the art that any block diagrams herein represent conceptual diagrams of illustrative circuitry embodying the principles of the invention. Similarly, it should be understood that any process sheet, flowchart, state transition diagram, pseudo-code, etc. represent various processes that can be substantially represented in a computer-readable medium and so executed by a computer or processor, whether explicitly or not Such a computer or processor is shown.

Furthermore, the following claims are hereby incorporated into the Detailed Description, with each claim standing on its own as a separate embodiment. Although each claim may stand on its own as a separate embodiment, it should be noted that although a dependent claim may in a claim refer to a specific combination with one or more other claims, other embodiments may also include the dependent claim Claims combine with the subject-matter of each other dependent claim. Unless it is stated that a particular combination is not intended, such a combination is suggested herein. Furthermore, it is intended that the features of a claim be included in any other independent claim, even if the claim is not directly dependent on the independent claim.

It should also be noted that the methods disclosed in the specification or claims may be implemented by an apparatus having means for performing each respective step of the methods.

List of reference signs

10 devices

12 One or more interfaces

14 Control Module

100 Mobile Transceivers

102 Another mobile transceiver

104 One or more active transceivers

110 Identifying Multiple Environment Models

120 Determining predicted future environmental models

120 Performing Time Series Forecasting

130 Predicting future service quality

135 Using Predicted Future Environment Models as Input to Machine Learning Models

140 Determining Quality of Service

145 Training Machine Learning Models

Claims (15)

1. A method for predicting future quality of service of a wireless communication link between a mobile transceiver (100) and another mobile transceiver (102), the method comprising: determining (110) a plurality of environment models for one or more active transceivers (104) in the environment of the mobile transceiver at a plurality of points in time; determining (120) a predicted future environment model for the one or more active transceivers at a future point in time using the time series forecast (125) for the plurality of environment models; and predicting (130) the future quality of service of the wireless communication link at future time points using a machine learning model, wherein the machine learning model is trained to provide information about the predicted quality of service for a given environment model, wherein the predicted future environment model is used as input to the machine learning model (135). 2. The method of claim 1, wherein the time series prediction is performed (125) based on a statistical fitting function or based on a temporal autocorrelation function. 3. The method according to any of the preceding claims, wherein the time series forecasting (125) is performed using another machine learning model. 4. The method according to any of the preceding claims, wherein the time series forecast is determined (125) such that the progress of the environmental model towards the predicted future environmental model is predicted. 5. The method of any preceding claim, wherein the time series forecast produces the predicted future environment model, wherein the future quality of service is predicted based on the predicted future environment model. 6. The method of any preceding claim, wherein the future quality of service of the wireless communication link is predicted (130) for at least two future points in time. 7. The method of claim 6, by determining (120) the predicted future environment model for the one or more active transceivers at the at least two future points in time, and using (135) the predicted future environment model of the one or more active transceivers at the at least two future points in time as input to the machine learning model, for the at least two future times The point predicts (130) the future quality of service for the wireless communication link. 8. The method of any preceding claim, comprising: determining (140) the quality of service of the wireless communication link at the plurality of points in time; and using the The plurality of environment models at the plurality of time points are used as training input and the quality of service of the wireless communication link at the corresponding plurality of time points is trained as the training output of the training of the machine learning model (145) The machine learning model. 9. The method of any preceding claim, wherein the machine learning model is trained (145) to implement a regression algorithm. 10. The method of any preceding claim, wherein the machine learning model is trained (145) to provide a probability distribution of the predicted quality of service given an environmental model. 11. The method of any preceding claim, wherein the one or more active transceivers are placed on a grid within the environment model, the grid comprising a plurality of contiguous cells, wherein the one or more active transceivers are aggregated in each cell within the grid. 12. The method of claim 11, wherein the grid is a circular grid. 13. The method of any preceding claim, wherein the predicted quality of service relates to at least one of inter-packet reception time, packet error rate, delay and data rate. 14. A computer program having program code for carrying out the method of any one of claims 1 to 13 when said computer program is executed on a computer, processor or programmable hardware component . 15. An apparatus (10) for predicting future quality of service of a wireless communication link between a mobile transceiver and another mobile transceiver, the apparatus comprising: one or more interfaces (15) for communicating in the mobile communication system; and A control module (14) configured to perform the method of any of claims 1 to 13.

Images